Project

meeseeker

0.01
No commit activity in last 3 years
No release in over 3 years
If you have multiple applications that need to perform actions as operations occur, `meeseeker` will allow your apps to each perform actions for specific operations without each app having to streaming the entire blockchain.
2005
2006
2007
2008
2009
2010
2011
2012
2013
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
 Dependencies

Development

>= 1.0.0, ~> 1.0
~> 5.10, >= 5.10.3
~> 0.6, >= 0.6.4
>= 1.0.0, ~> 1.0
>= 0.22.0, ~> 0.22
~> 0.11, >= 0.11.3
~> 12.3, >= 12.3.1
~> 0.15, >= 0.15.1

Runtime

>= 1.0.0, ~> 1.0.0
~> 2.7, >= 2.7.6
~> 0.5, >= 0.5.5
>= 4.1.0, ~> 4.1
~> 0.9, >= 0.9.4
 Project Readme

meeseeker

Redis based block follower is an efficient way for multiple apps to stream the Hive Blockchain.

Build Status

If you have multiple applications that need to perform actions as operations occur, meeseeker will allow your apps to each perform actions for specific operations without each app having to stream the entire blockchain.

In a nutshell: The overarching intent here is to provide a "live view" of the blockchain, not store the entire blockchain. Apps can attach to your redis source and ask, "What just happened?"

Purpose

Although Meeseeker tracks all operations, it is only intended to provide other applications signals that those operations have happened. It is not intended to provide cryptographically verifiable events.

Possible uses:

  • Notifications of events, suitable for push to mobile devices or web browsers.
  • Invoke periodic updates on a threshold.
  • Light-weight bots that only care about a limit set of operations, reducing the number of API calls.

Why Redis?

Redis is a persistent key-value database, with built-in net interface. See: https://redis.io/

It allows for quick storage and lookup of operations by key as well as the ability to automatically expire keys that are no longer needed.

Installation

First, install redis:

On linux:

sudo apt install redis-server

On macOS:

brew install redis

Next, install ruby. One way to do this is install rvm. Once ruby is installed, install meeseeker with the gem command:

gem install meeseeker

This installs meeseeker as a command available to the OS, e.g.:

meeseeker help

To do the actual sync to your local redis source (defaults assume redis://127.0.0.1:6379/0):

meeseeker sync

To specify an alternative redis source:

MEESEEKER_REDIS_URL=redis://:p4ssw0rd@10.0.1.1:6380/15 meeseeker sync

You can also specify am alternative Hive node:

MEESEEKER_NODE_URL=http://anyx.io meeseeker sync

You can also specify a Steem node instead of Hive (if that's your thing):

MEESEEKER_NODE_URL=https://api.steemit.com meeseeker sync[steem]

Or, you can have meeseeker automatically use random Hive nodes:

MEESEEKER_NODE_URL=shuffle meeseeker sync

To sync from the head block instead of the last irreversible block:

MEESEEKER_STREAM_MODE=head meeseeker sync

To ignore virtual operations (useful if the node doesn't enable get_ops_in_blocks or if you want to sync from the head block):

MEESEEKER_INCLUDE_VIRTUAL=false meeseeker sync

Normally, block headers are added to the hive:block channel. This requires one additional API call for each block. If you don't need block headers, you can configure the hive:block channel to only publish with the block_num:

MEESEEKER_INCLUDE_BLOCK_HEADER=false meeseeker sync

Normally, keys stay on redis for 24 hours. If you want to change this behavior, use MEESEEKER_EXPIRE_KEYS and specify the new value in seconds, for example:

MEESEEKER_EXPIRE_KEYS=10 meeseeker sync

If you never want the keys to expire (not recommended), set MEESEEKER_EXPIRE_KEYS to -1:

MEESEEKER_EXPIRE_KEYS=-1 meeseeker sync

Normally, sync will create keys until it uses up all available memory. If you would like to only sync a certain number of keys, then sleep until those keys expire so it can pick up where it left off, set MEESEEKER_MAX_KEYS to a positive value:

MEESEEKER_MAX_KEYS=99 meeseeker sync

Usage

When meeseeker sync starts for the first time, it initializes from the last irreversible block number. If the sync is interrupted, it will resume from the last block sync'd unless that block is older than MEESEEKER_EXPIRE_KEYS in which case it will skip to the last irreversible block number.

Using SUBSCRIBE

For redis-cli, please see: https://redis.io/topics/pubsub

Sync

When running meeseeker sync, the following channels are available:

  • hive:block
  • hive:transaction
  • hive:op:vote
  • hive:op:comment
  • hive:op:comment_options
  • hive:op:whatever (replace "whatever" with the op you want)
  • hive:op:custom_json:whatever (if enabled, replace "whatever" with the custom_json.id you want)

As mentioned in the first whatever example, for ops, all operation types can be subscribed to as channels, including virtual operations, if enabled.

In the second whatever example, for custom_json.id, if you want to subscribe to the follow channel, use hive:op:custom_json:follow. Or if you want to subscribe to the sm_team_reveal channel, use hive:op:custom_json:sm_team_reveal. The custom_json.id channels are not enabled by default. To enable it, set the MEESEEKER_PUBLISH_OP_CUSTOM_ID to true (see example below).

For example, from redis-cli, if we wanted to stream block numbers:

$ redis-cli
127.0.0.1:6379> subscribe hive:block
Reading messages... (press Ctrl-C to quit)
1) "subscribe"
2) "hive:block"
3) (integer) 1
1) "message"
2) "hive:block"
3) "{\"block_num\":29861068,\"previous\":\"01c7a4cb4424b4dc0cb0cc72fd36b1644f8aeba5\",\"timestamp\":\"2019-01-28T20:55:03\",\"witness\":\"ausbitbank\",\"transaction_merkle_root\":\"a318bb82625bd78af8d8b506ccd4f53116372c8e\",\"extensions\":[]}"
1) "message"
2) "hive:block"
3) "{\"block_num\":29861069,\"previous\":\"01c7a4cc1bed060876cab57476846a91568a9f8a\",\"timestamp\":\"2019-01-28T20:55:06\",\"witness\":\"followbtcnews\",\"transaction_merkle_root\":\"834e05d40b9666e5ef50deb9f368c63070c0105b\",\"extensions\":[]}"
1) "message"
2) "hive:block"
3) "{\"block_num\":29861070,\"previous\":\"01c7a4cd3bbf872895654765faa4409a8e770e91\",\"timestamp\":\"2019-01-28T20:55:09\",\"witness\":\"timcliff\",\"transaction_merkle_root\":\"b2366ce9134d627e00423b28d33cc57f1e6e453f\",\"extensions\":[]}"

In addition to general op channels, there's an additional channel for custom_json.id. This option must be enabled:

MEESEEKER_PUBLISH_OP_CUSTOM_ID=true meeseeker sync

Which allows subscription to specific id patterns:

$ redis-cli
127.0.0.1:6379> subscribe hive:op:custom_json:sm_team_reveal
Reading messages... (press Ctrl-C to quit)
1) "subscribe"
2) "hive:op:custom_json:sm_team_reveal"
3) (integer) 1
1) "message"
2) "hive:op:custom_json:sm_team_reveal"
3) "{\"key\":\"hive:29890790:bcfa68d9be10b3587d81039b85fd0536ddeddffb:0:custom_json\"}"
1) "message"
2) "hive:op:custom_json:sm_team_reveal"
3) "{\"key\":\"hive:29890792:3f3b921ec6706bcd259f5cc6ac922dc59bbe2de5:0:custom_json\"}"
1) "message"
2) "hive:op:custom_json:sm_team_reveal"
3) "{\"key\":\"hive:29890792:4ceca16dd114b1851140086a82a5fb3a6eb6ec42:0:custom_json\"}"
1) "message"
2) "hive:op:custom_json:sm_team_reveal"
3) "{\"key\":\"hive:29890792:00930eff76b3f0af8ed7215e88cf351cc671490b:0:custom_json\"}"
1) "message"
2) "hive:op:custom_json:sm_team_reveal"
3) "{\"key\":\"hive:29890799:01483bd252ccadb05f546051bb20a4ba9afea243:0:custom_json\"}"

A ruby application can subscribe to a channel as well, using the redis gem:

require 'redis'

url = 'redis://127.0.0.1:6379/0'
ctx = Redis.new(url: url)

Redis.new(url: url).subscribe('hive:op:comment') do |on|
  on.message do |channel, message|
    payload = JSON[message]
    comment = JSON[ctx.get(payload['key'])]
    
    puts comment['value']
  end
end

Many other clients are supported: https://redis.io/clients

Witness Schedule

When running meeseeker witness:schedule, the hive:witness:schedule channel is available. This is offered as a separate command because most applications don't need to worry about this level of blockchain logistics.

For example, from redis-cli, if we wanted to subscribe to the witness schedule:

$ redis-cli
127.0.0.1:6379> subscribe hive:witness:schedule
Reading messages... (press Ctrl-C to quit)
1) "subscribe"
2) "hive:witness:schedule"
3) (integer) 1
1) "message"
2) "hive:witness:schedule"
3) "{\"id\":0,\"current_virtual_time\":\"415293532210075480213212125\",\"next_shuffle_block_num\":30035208,\"current_shuffled_witnesses\":[\"thecryptodrive\",\"timcliff\",\"utopian-io\",\"themarkymark\",\"aggroed\",\"smooth.witness\",\"someguy123\",\"gtg\",\"followbtcnews\",\"yabapmatt\",\"therealwolf\",\"ausbitbank\",\"curie\",\"clayop\",\"drakos\",\"blocktrades\",\"good-karma\",\"roelandp\",\"lukestokes.mhth\",\"liondani\",\"anyx\"],\"num_scheduled_witnesses\":21,\"elected_weight\":1,\"timeshare_weight\":5,\"miner_weight\":1,\"witness_pay_normalization_factor\":25,\"median_props\":{\"account_creation_fee\":{\"amount\":\"3000\",\"precision\":3,\"nai\":\"@@000000021\"},\"maximum_block_size\":65536,\"sbd_interest_rate\":0,\"account_subsidy_budget\":797,\"account_subsidy_decay\":347321},\"majority_version\":\"0.20.8\",\"max_voted_witnesses\":20,\"max_miner_witnesses\":0,\"max_runner_witnesses\":1,\"hardfork_required_witnesses\":17,\"account_subsidy_rd\":{\"resource_unit\":10000,\"budget_per_time_unit\":797,\"pool_eq\":157691079,\"max_pool_size\":157691079,\"decay_params\":{\"decay_per_time_unit\":347321,\"decay_per_time_unit_denom_shift\":36},\"min_decay\":0},\"account_subsidy_witness_rd\":{\"resource_unit\":10000,\"budget_per_time_unit\":996,\"pool_eq\":9384019,\"max_pool_size\":9384019,\"decay_params\":{\"decay_per_time_unit\":7293741,\"decay_per_time_unit_denom_shift\":36},\"min_decay\":257},\"min_witness_account_subsidy_decay\":0}"

Using SCAN

From the redis manual:

Since these commands allow for incremental iteration, returning only a small number of elements per call, they can be used in production without the downside of commands like KEYS or SMEMBERS that may block the server for a long time (even several seconds) when called against big collections of keys or elements.

However while blocking commands like SMEMBERS are able to provide all the elements that are part of a Set in a given moment, The SCAN family of commands only offer limited guarantees about the returned elements since the collection that we incrementally iterate can change during the iteration process.

See: https://redis.io/commands/scan

Keep in mind that SCAN requires pagination to get a complete result. Redis implements pagination using a cursor based iterator.

See: https://redis.io/commands/scan#scan-basic-usage

Once your sync has started, you can begin doing queries against redis, for example, in the redis-cli:

redis-cli --scan --pattern 'hive:*:vote'

This returns the keys, for example:

hive:29811083:7fd2ea1c73e6cc08ab6e24cf68e67ff19a05896a:0:vote
hive:29811085:091c3df76322ec7f0dc51a6ed526ff9a9f69869e:0:vote
hive:29811085:24bfc199501779b6c2be2370fab1785f58062c5a:0:vote
hive:29811086:36761db678fe89df48d2c5d11a23cdafe57b2476:0:vote
hive:29811085:f904ac2e5e338263b03b640a4d1ff2d5fd01169e:0:vote
hive:29811085:44036fde09f20d91afda8fc2072b383935c0b615:0:vote
hive:29811086:570abf0fbeeeb0bb5c1e26281f0acb1daf175c39:0:vote
hive:29811083:e3ee518c4958a10f0d0c5ed39e3dc736048e8ec7:0:vote
hive:29811083:e06be9ade6758df59e179160b749d1ace3508044:0:vote

To get the actual vote operation for a particular key, use:

redis-cli get hive:29811085:f904ac2e5e338263b03b640a4d1ff2d5fd01169e:0:vote

If, on the other hand, you want custom_json only:

redis-cli --scan --pattern 'hive:*:custom_json'

This only returns the related keys, for example:

hive:29811084:43f1e1a367b97ea4e05fbd3a80a42146d97121a2:0:custom_json
hive:29811085:5795ff73234d64a11c1fb78edcae6f5570409d8e:0:custom_json
hive:29811083:2d6635a093243ef7a779f31a01adafe6db8c53c9:0:custom_json
hive:29811086:31ecb9c85e9eabd7ca2460fdb4f3ce4a7ca6ec32:0:custom_json
hive:29811083:7fbbde120aef339511f5af1a499f62464fbf4118:0:custom_json
hive:29811083:04a6ddc83a63d024b90ca13996101b83519ba8f5:0:custom_json

To get the actual custom json operation for a particular key, use:

redis-cli get hive:29811083:7fbbde120aef339511f5af1a499f62464fbf4118:0:custom_json

To get all transactions for a particular block number:

redis-cli --scan --pattern 'hive:29811085:*'

Or to get all ops for a particular transaction:

redis-cli --scan --pattern 'hive:*:31ecb9c85e9eabd7ca2460fdb4f3ce4a7ca6ec32:*'

Hive Engine Support

As of v0.0.6, meeseeker can also follow the Hive Engine side-chain. This is optional and requires a separate process.

To sync Hive Engine to your local redis source (also defaults to redis://127.0.0.1:6379/0):

meeseeker sync hive_engine

When running meeseeker sync hive_engine, the following channels are available:

  • hive_engine:block
  • hive_engine:transaction
  • hive_engine:virtual_transaction
  • hive_engine:contract
  • hive_engine:contract:deploy
  • hive_engine:contract:update
  • hive_engine:market
  • hive_engine:market:buy
  • hive_engine:market:cancel
  • hive_engine:market:sell
  • hive_engine:sscstore
  • hive_engine:sscstore:buy
  • hive_engine:steempegged
  • hive_engine:steempegged:buy
  • hive_engine:steempegged:removeWithdrawal
  • hive_engine:steempegged:withdraw
  • hive_engine:tokens
  • hive_engine:tokens:checkPendingUnstake
  • hive_engine:tokens:create
  • hive_engine:tokens:enableStaking
  • hive_engine:tokens:issue
  • hive_engine:tokens:transfer
  • hive_engine:tokens:transferOwnership
  • hive_engine:tokens:unstake
  • hive_engine:tokens:updateMetadata
  • hive_engine:tokens:updateParams
  • hive_engine:tokens:updateUrl

The above "channel/action" patterns are the ones that are known that the time of writing. In addition, if a new contract is added or updated, meeseeker will automatically publish to these corresponding channels as they appear, without needing to update or even restart meeseeker.

See main section on Using SUBSCRIBE.

Once your HiveEngine sync has started, you can begin doing queries against redis, for example, in the redis-cli:

redis-cli --scan --pattern 'hive_engine:*:tokens:transfer'

This returns the keys, for example:

hive_engine:18000:d414373db84e6a642f289641ea1433fda22b8a4d:0:tokens:transfer
hive_engine:18004:c9e06c8449d2d04b4a0a31ec7b80d2f62009a5f0:0:tokens:transfer
hive_engine:17994:faf097391760ad896b19d5854e2822f62dee284b:0:tokens:transfer

See main section on Using SCAN.

Docker

This will launch meeseeker in a docker container, so you can immediately attach to it on port 6380.

docker run -d -p 6380:6379 inertia/meeseeker:latest
redis-cli -p 6380

You can also pass any of the environment variables meeseeker accepts. For example, this will launch meeseeker with custom_json.id channels enabled, but only keeps ops around for 5 minutes:

docker run \
  --env MEESEEKER_PUBLISH_OP_CUSTOM_ID=true \
  --env MEESEEKER_EXPIRE_KEYS=300 \
  -d -p 6380:6379 inertia/meeseeker:latest

Also see: https://hub.docker.com/r/inertia/meeseeker/


See some of my previous Ruby How To posts in: #radiator #ruby

Get in touch!

If you're using Radiator, I'd love to hear from you. Drop me a line and tell me what you think! I'm @inertia on Hive.

License

I don't believe in intellectual "property". If you do, consider Radiator as licensed under a Creative Commons CC0 License.